Cardiovascular Disease Risk Missed in Many HIV-Positive Veterans Despite Better Risk-Assessment Tool
October 21, 2014
A retrospective analysis using the 2013 American College of Cardiology/American Heart Association (ACC/AHA) Guidelines and Pooled Cohort Equations (PCEs) for acute myocardial infarction (AMI) and cardiovascular disease (CVD) risk found that the PCE scores were a much better predictor of 10-year CVD risk than the Framingham risk score in a cohort of more than 15,000 HIV-positive veterans on antiretroviral therapy, according to a presentation at ICAAC 2014, made by Henning Drechsler, M.D., of the University of Texas Southwestern Medical Center.
However, the PCE scores still greatly underestimated the 10-year CVD risk in the veterans living with HIV, and adding other covariates did not significantly affect the model's performance. While use of the 2013 ACC/AHA Guidelines was able to identify the majority of those who would go on to have an AMI or CVD, 40% of events were still missed.
Even so, the analysis found that application of the 2013 ACC/AHA Guidelines dramatically increased (by four to fivefold) the number of those who qualified for statins. Overall, as many as 50% of veterans living with HIV in this cohort qualified.
CVD has been responsible for an increasing proportion of the mortality in people living with HIV in the antiretroviral therapy era. The relative risk of CVD in people living with HIV is 50% to 100% higher than in the HIV-negative population -- and higher in people on antiretroviral therapy. The HIV-associated risk of AMI is 50% higher than can be explained by other concurrent risk factors in the population. The reasons are insufficiently understood -- it could be due to HIV, higher rates of comorbidities, or as a paper presented at CROI 2013 by Tenorio et al suggested, chronic inflammation or immune reconstitution (note, this paper is not available online).
Despite the greater risk, clinicians lack a good HIV-specific CVD risk assessment tool. As reported at IDWeek last year, two possible tools -- both the Framingham as well as the HIV-specific D:A:D risk score -- disappointed in terms of absolute performance.
The Framingham risk scores were the tool recommended by the 2004 Adult Treatment Panel (ATP) III lipid guidelines to determine who was at risk of CVD and in need of statins. However, in 2013, the joint ACC/AHA coronary heart disease prevention guideline had its focus widened from coronary heart disease to cerebro and cardiovascular disease risk. In addition, the recommended risk calculation tool is now the PCE, which, in contrast to the Framingham risk equations, also incorporates diabetes and African-American race into its equation. Finally, the guidelines' focus is no longer on shifting low-density lipoprotein cholesterol (LDL) targets (as per the 2004 ATP III lipid guidelines) but on whether a patient requires a high or moderate intensity statin.
So Drechsler and colleagues decided to conduct a retrospective analysis of the local Veterans Affairs (VA) cohort, to see how the two scores (the PCE and the Framingham) performed in absolute and in relative terms in an HIV-positive population, and to see whether the addition of established HIV-specific covariates would improve the equations' performance. They also wanted to assess the proportion of patients who would meet statin indications under the old lipid-lowering guidelines and whether the ACC/AHA indication for taking a statin after starting antiretroviral therapy could correctly identify patients who would suffer subsequent AMI/CVD events.
The study included any veteran living with HIV from the cohort who had been on antiretroviral therapy for at least 15 months without a qualifying endpoint during that time, and if there were enough information available to calculate a risk score. Score calculation required at least one total cholesterol (TC) value, one high-density lipoprotein cholesterol (HDL), and one blood pressure (BP) measurement in the first 15 months of going onto antiretroviral therapy. Antihypertensive treatment was assumed if a veteran was in possession of an antihypertensive at the day of BP measurement; and smoking was assumed if a specific International Classification of Diseases-9 (ICD-9) code was used in their charts or a varenicline (brand name Chantix) prescription was filled before month 15.
In the analysis, qualifying events included both AMI, as defined by an ICD-9 code of 410.x (402 events); as well as ischemic strokes, also identified by ICD-9 codes: 431.x, 432.9, 433.x1, 434.x, 437.1, 438.x (450 events). The combined endpoint was whichever of the two above occurred first: 801 (note, some patients experienced both events).
The study's first objective was to determine which model predicts events better. To do this, the researchers looked at the average Framingham score in each of the PCE risk strata and compared it with the observed AMI incidence rates. What they found is that for all but the lowest risk strata, the Framingham risk score overpredicted AMI incidence (this has been noted before in the D:A:D cohort). In contrast, PCE significantly underpredicted CVD events with the exception of the highest risk group.
The Framingham risk scores and PCE were then compared head-to-head using the guideline relevant cutoffs (for high risk) of 7.5% and 10%. Unsurprisingly, there was a high hazard ratio (HR) for more CVD in individuals who were at high risk using both scores. However, the veterans who were at high risk by the Framingham, but not the PCE, did not have a subsequently elevated event rate for cardiovascular disease, while there was a significantly elevated risk for those individuals with a high PCE score and a low Framingham risk score -- and the same was also true for AMI events (HR 2.06 [1.43-2.98]).
The researchers also explored whether they could improve the multivariate model by including other covariates (a number of which were specific to this population of people living with HIV), such as
Although several of these are indeed significant predictors of CVD, their inclusion in the multivariate models did not improve its performance.
The researchers' third objective was to see how many of the veterans would need statins based on the 2013 ACC/AHA Guidelines versus the old ATP III criteria (this analysis excluded those veterans who were not already on statins). Under the old criteria, only 1,611 qualified for statins. Under the 2013 guidelines, the number was four to five times higher: 7,860 -- roughly half of the entire cohort.
But does the indication for statin improve the identification of those with subsequent cardiovascular events? Looking at the event numbers that occurred by the ACC/AHA or ATP III categorization -- a large number were only predicted by the ACC/AHA indication, but there were also some false positives. So researchers calculated the net reclassification improvement (NRI) as the proportional improvement in true positives (36.7%) minus the proportional worsening in false positives (25.7%), which yielded a highly significant NRI of 11% for combined cardiovascular events, and an NRI for AMIs of 9.3%, which was also significant. Even so, 40% of the individuals with events would not have qualified for statins using either score.
There were a few limitations to the study. The chart review didn't allow for event ascertainment, but the researchers did perform sensitivity analyses where they required additional laboratory evidence for myocardial infarctions and brain imaging evidence for strokes -- the results, according to Drechsler were very similar.
In addition, an assessment of the risks of currently smoking status wasn't possible because the data only captured smoking status as "having ever smoked."
More disappointing is that the data are not generalizable to women living with HIV, because there were so few in the cohort.
Nevertheless, the exploration of these guidelines would be a positive step in the right direction.
"If we want to be serious about cardiovascular prevention we need to put fifty percent of our veterans on statins," Drechsler concluded.
Theo Smart is an HIV activist and medical writer with more than 20 years of experience. You can follow him on Twitter @theosmart.
Copyright © 2014 Remedy Health Media, LLC. All rights reserved.
This article was provided by TheBodyPRO. It is a part of the publication The 54th Interscience Conference on Antimicrobial Agents and Chemotherapy (ICAAC 2014).
Add Your Comment:
(Please note: Your name and comment will be public, and may even show up in
Internet search results. Be careful when providing personal information! Before
adding your comment, please read TheBody.com's Comment Policy.)
The content on this page is free of advertiser influence and was produced by our editorial team. See our content and advertising policies.